Before every client review, a relationship manager assembles a pack: holdings, recent activity, product fit, what was promised last time, and what compliance will allow them to say. It happens the night before, and its quality depends on which RM you get.
That is a workflow problem, and it is one AI is well suited to, as long as the RM stays in charge.
What the draft pack contains
The pack is drafted automatically when the review is booked, from systems the bank already runs:
- Holdings and recent activity, from the portfolio system
- What was discussed and promised last time, from CRM notes
- Signals worth raising: changes in the client's behaviour, maturing positions, idle cash
- Talking points, each with a source the RM can click through to
- Flags on any line compliance would question, such as product suitability or performance claims
Dataiku's guide describes a similar pattern for wealth managers: pulling together trading, CRM and market data, flagging shifts in client behaviour and suggesting a next action. The difference that matters in a bank is what happens next.
Where the RM decides
The RM edits and owns the pack. Nothing goes to a client without the RM, and AI never makes a recommendation to a client. Lines flagged for compliance are reviewed before the meeting. Follow-up actions agreed in the meeting are logged and diarised in CRM.
What compliance sees
Compliance sees fewer packs, earlier. Instead of reviewing whatever an RM wrote the night before, the reviewer sees only the flagged lines, with their sources, in time to act. Every version of the pack is kept, so the one actually used in the meeting can be found a year later.
In Camunda's September 2026 survey, compliance and risk operations was the process organisations most wanted to redesign, named by 68%. The client pack is one place where that redesign is visible to the front office as well as to compliance.
What stays inside the bank
The workflow runs inside the bank's environment and CRM, with a model provider that IT and risk have approved. No client data goes into public AI tools, and every action is logged.
What to measure
- Prep time per review, from an RM time diary before and after
- Reviews per RM per month, from CRM
- Compliance edits per pack, from the review log
- Follow-ups completed on time, from CRM tasks
Where to start
Start with one team of RMs and their compliance reviewer, in the same room. Map how a pack is built today, agree the baseline, and write down what the AI may never do. Our two-week diagnostic ends with exactly that, and a ranked shortlist of what to build first.